
Sampling Design
Terrestrial sampling options
In this file, the goal is to:
- Examine the environmental coverage available within focal sites
- Attempt to subdivide those focal sites into clusters for representative sampling
Sampling options
Focal sites
Focal study sites have been selected on the basis of the following criteria:
- occur in the domain
- occur within a national park
- occur within 1 km of a road
- have a mean NDVI over the last year of > 0.2
- have a slope < 30 degrees
- are within an estimated hiking time of 2 hours from the road
Clustering
To better stratify our sampling, we’ll use k-means clustering to divide up the cells within the domain into clusters.
The variables going into the clustering are:
Distance from water (including either rivers or wetlands, but not including the ocean)- Drought (the number of droughts since 2000, quantified as the number SPEI values less than -1 since 2000)
- Soil Depth (from ISDASOIL, depth to bedrock, with a maximum values of 2 meters)
- Annual Mean Temperature
- Annual Precipitation
- Precipitation Seasonality (Coefficient of Variation)
Here I’m using k-means clustering to divide the data into 20 clusters. The number 20 was chosen entirely arbitrarily.
Visualizing our clusters
Geographic space


Environmental space



Vegetation Types

Interactive map
The map below is interactive and allows zooming, moving, and toggling layers on/off. The cluster polygons are available here.